The industry is increasingly attracted by Additive Manufacturing (AM) processes thanks to their possibility of overcoming many limitations of conventional technologies. With AM, it is possible to manufacture parts with very complex geometries, including internal channels and lattices. These features have made AM interesting in all those sectors where high customization or high-performance resistant lightweight parts are required, such as the aerospace, medical and automotive ones. Due to the complexity of the physical phenomena involved and still evolving maturity of the technology, AM processes may lack adequate repeatability and capability with respect to very stringent quality requirements imposed by the industry. On the one side, the long durations of manufacturing of AM processes make inconvenient to wait until the completion of the final part to perform inspections through expensive and time-consuming Non-Destructive Techniques (NDT). On the other side, the layerwise nature of the process is suitable to perform real time monitoring of the process on a layer by layer basis. Therefore, the development of a real time monitoring tool would be a relevant step forward in providing AM processes the robustness to be competitive in industrial applications. In-situ monitoring in AM is performed through vision systems and high-speed videos that capture fast transient phenomena, which are proxies of the stability of the process. This generates streams of big data, which are very difficult to manage, especially in order to obtain capability of detecting out-of-control events in real time. This thesis work is aimed at studying a method for reducing the dimensionality of in-situ acquired data, by condensing the information contained in streams of 2D images into 1D profiles. Thanks to this approach, based on the exploitation of Ripley’s K-function, it is possible to manage in-situ streams of big data and develop monitoring techniques through statistical analysis of 1D profiles. The developed method was applied to high-speed videos in Selective Laser Melting (SLM), with focus on the by-products of the process. Previous studies showed that the by-products of the laser beam and material interactions, i.e., spatters, may be used as proxies of the process stability in SLM, and recent research efforts have been carried out to find spatter-related descriptors suitable for in-situ monitoring. This thesis work shows a further way of exploitation of the K-function as a tool for modelling internal porosity detected in X-Ray Computed Tomography (CT) images.
Il mondo dell’industria è sempre più attratto dai processi di manifattura additiva per la possibilità di superare molte delle limitazioni imposte dai processi convenzionali. Con i processi additivi, è possibile produrre parti con geometrie molto complesse, inclusi canali e strutture reticolari interne. Queste caratteristiche hanno reso i processi additivi particolarmente interessanti in tutti quei contesti, come quelli aerospaziale, medico e automobilistico, dove sono richiesti una alta personalizzazione della parte, oppure parti molto performanti, che siano allo stesso tempo leggere e resistenti. A causa della complessità dei fenomeni fisici interessati e della maturità della tecnologia, ancora in evoluzione, i processi additivi potrebbero essere carenti di una adeguata ripetibilità e competenze per soddisfare i requisiti di qualità molto stringenti imposti dall’industria. Da un lato, i lunghi tempi di produzione rendono sconveniente aspettare che la parte sia completata per procedere con Prove Non Distruttive, costose sia in termini economici che temporali, finalizzate ad attestarne la qualità. Dall’altro, la natura stessa dei processi additivi si presta ad un tipo di monitoraggio in tempo reale di ogni strato depositato. Dunque, lo sviluppo di uno strumento di monitoraggio in tempo reale costituirebbe un significativo passo in avanti nel fornire ai processi additivi la robustezza di cui necessitano per essere competitivi in applicazioni industriali. Il monitoraggio in-situ dei processi additivi viene implementato attraverso sistemi di visione e video ad alta velocità, in grado di catturare fenomeni transitori che siano indicatori della stabilità del processo. Questo genera flussi di dati di grandi dimensioni, difficili da gestire, specialmente nell’ottica di rilevare anomalie in tempo reale. Questo lavoro di tesi si propone di studiare un metodo per ridurre la dimensionalità dei dati acquisiti in-situ, sintetizzando l’informazione contenuta in flussi di immagini 2D in profili 1D. Grazie a questo approccio, basato sull’utilizzo della funzione K di Ripley, è possibile gestire flussi di dati di grandi dimensioni e sviluppare tecniche di monitoraggio attraverso l’analisi statistica dei profili 1D estrapolati. Il metodo sviluppato è stato applicato a video ad alta velocità del processo di Fusione Laser Selettiva, con particolare riferimento agli effetti collaterali del processo. Studi precedenti hanno mostrato che questi effetti collaterali, nella fattispecie le scintille, potrebbero essere usati come indicatori della stabilità del processo, e recentemente sono stati portati avanti tentativi di ricerca volti a trovare descrittori, legati alle scintille, che possano essere adatti al monitoraggio in-situ. Questo lavoro di tesi presenta un ulteriore possibile utilizzo della funzione K per modellare la porosità interna in immagini acquisite tramite Tomografia Computerizzata ai raggi-X.
In-situ monitoring of SLM process through Ripley's K-function applied to spatters spatial distribution
MARCONI, NICOLO'
2018/2019
Abstract
The industry is increasingly attracted by Additive Manufacturing (AM) processes thanks to their possibility of overcoming many limitations of conventional technologies. With AM, it is possible to manufacture parts with very complex geometries, including internal channels and lattices. These features have made AM interesting in all those sectors where high customization or high-performance resistant lightweight parts are required, such as the aerospace, medical and automotive ones. Due to the complexity of the physical phenomena involved and still evolving maturity of the technology, AM processes may lack adequate repeatability and capability with respect to very stringent quality requirements imposed by the industry. On the one side, the long durations of manufacturing of AM processes make inconvenient to wait until the completion of the final part to perform inspections through expensive and time-consuming Non-Destructive Techniques (NDT). On the other side, the layerwise nature of the process is suitable to perform real time monitoring of the process on a layer by layer basis. Therefore, the development of a real time monitoring tool would be a relevant step forward in providing AM processes the robustness to be competitive in industrial applications. In-situ monitoring in AM is performed through vision systems and high-speed videos that capture fast transient phenomena, which are proxies of the stability of the process. This generates streams of big data, which are very difficult to manage, especially in order to obtain capability of detecting out-of-control events in real time. This thesis work is aimed at studying a method for reducing the dimensionality of in-situ acquired data, by condensing the information contained in streams of 2D images into 1D profiles. Thanks to this approach, based on the exploitation of Ripley’s K-function, it is possible to manage in-situ streams of big data and develop monitoring techniques through statistical analysis of 1D profiles. The developed method was applied to high-speed videos in Selective Laser Melting (SLM), with focus on the by-products of the process. Previous studies showed that the by-products of the laser beam and material interactions, i.e., spatters, may be used as proxies of the process stability in SLM, and recent research efforts have been carried out to find spatter-related descriptors suitable for in-situ monitoring. This thesis work shows a further way of exploitation of the K-function as a tool for modelling internal porosity detected in X-Ray Computed Tomography (CT) images.| File | Dimensione | Formato | |
|---|---|---|---|
|
Tesi_Marconi.pdf
non accessibile
Descrizione: Testo della tesi
Dimensione
5.64 MB
Formato
Adobe PDF
|
5.64 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/148800